0% found this document useful (0 votes)
111 views27 pages

Linguistic Relativity

This document discusses the concept of linguistic relativity and whether language shapes thought. It outlines the hypothesis that each language embodies a particular worldview and impacts how speakers of that language understand reality. The strongest version of the hypothesis is that language differences necessarily cause thought differences. Weaker versions argue that language and thought are simply different across language communities. The document examines whether aspects of language grammar, such as classifications and references, can influence thought and worldviews. It uses examples of translations between languages to illustrate significant differences in how reality is described.

Uploaded by

Raul Costa
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
111 views27 pages

Linguistic Relativity

This document discusses the concept of linguistic relativity and whether language shapes thought. It outlines the hypothesis that each language embodies a particular worldview and impacts how speakers of that language understand reality. The strongest version of the hypothesis is that language differences necessarily cause thought differences. Weaker versions argue that language and thought are simply different across language communities. The document examines whether aspects of language grammar, such as classifications and references, can influence thought and worldviews. It uses examples of translations between languages to illustrate significant differences in how reality is described.

Uploaded by

Raul Costa
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 27

Institute of General Semantics

LINGUISTIC RELATIVITY
Author(s): Eleanor Rosch
Source: ETC: A Review of General Semantics, Vol. 44, No. 3 (Fall 1987), pp. 254-279
Published by: Institute of General Semantics
Stable URL: http://www.jstor.org/stable/42579365
Accessed: 27-12-2015 14:23 UTC

REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/42579365?seq=1&cid=pdf-reference#references_tab_contents

You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact [email protected].

Institute of General Semantics is collaborating with JSTOR to digitize, preserve and extend access to ETC: A Review of
General Semantics.

http://www.jstor.org

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
LINGUISTIC
EleanorRosch* RELATIVITY*

we "trapped" byour language into holdinga particular"world


Are view?"Can we neverreallyunderstandor communicatewithspeakers
ofa languagequitedifferent fromourownbecauseeach languagehas molded
thethoughtofitspeople intomutuallyincomprehensible worldviews?Can
we neverget"beyond"languagetoexperiencetheworld"directly"? Such issues
develop from an extreme form of a position sometimes known as "theWhorfian
hypothesis," after the earlytwentieth centurylinguistBenjamin Lee Whorf,
and called, moregenerally, the hypothesisof "linguisticrelativity."
Accordingto linguisticrelativity, it is naiveto thinkthatwhenwe learna
"foreign" we
language, simply learn a newvocabularyto namethesameobjects
and a new grammarto expressthesame relationsbetweenobjectsas existin
ourownlanguage.Rather,"thebackgroundlinguisticsystem... ofeach lan-
guageis notmerelya reproducinginstrument forvoicingideas but ratheris
itselftheshaperofideas. . . . We dissectnaturealong lineslaid downbyour
nativelanguage.The categoriesand typesthatwe isolatefromtheworldof
phenomenawe do notfindtherebecausetheystareeveryobserverin thefece;
on thecontrary, theworldis presentedin a kaleidoscopicfluxofimpressions
whichhas tobe organizedby our minds- and thismeans ... bythelinguis-
tic systemin our minds. (1)
When manyof us firstcame in contactwiththe Whorfianhypothesis, it
seemednotonlytruebut profoundly true.We feltwe could look inwardand
see our comprehensionof the world molded by languagejust as we could
"watch"as our personalitieswereirrevocably shapedbysocietyand upbring-
ing. But profound and ineffable truths are not,in thatform,subjectto scien-
tificinvestigation. Is linguisticrelativity an empirical"theory"?If so,itmust
be possibletoderivefromitconcrete statements aboutspecificrelationsofactual
languages to the thought of the people thatspeak them; and these statements

*Eleanor
Roschisa professor
ofpsychology
attheUniversity
ofCalifornia,
Berkeley.
fThisarticle
wasreprinted
fromHuman Theoretical
Communication: of
, bypermission
Explorations
Erlbaum
Lawrence Associates.

254

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
Linguistic Relativity 255

mustbe ofa typewhichcan be judgedtrueor falsebycomparingthemtofacts


about thoseactual languagesand thoughts.
Therearea numberofimportant withintheWhorfianposition
distinctions
empiricalimplications.Relatively"weak"or "strong"
whichlead to differing
claims maybe assertedabout the role of languagein thought:at the weak
extremeis thesimpleclaimthatbothlanguagesand thoughtsaredifferent in
diffèrent
languagecommunities, whilethestrongversionis thatlanguagediffer-
encesnecessarilycause(are necessaryand sufficientconditionsfor)thought
differences.The strongerclaimis sometimescalled LinguisticDeterminism
todistinguish Perhapsevenmore
itfromthelessspecificLinguisticRelativity.
important, aredistinctions
operationally, amongtheaspects(or units)oflan-
guageconsideredas potentialinfluencersofthought.Arewe focusingon overt
or covertunitsin language?On basic grammaticaldivisions?On grammati-
cal formclassesin general?On classifications inherentin lan-
and references
guagevocabularies?We will considerthesepossibilitiesin turn.

COVERT LINGUISTIC CLASSIFICATIONS

Language as Metaphysics
The strongest and mostinclusiveformoftheWhorfianhypothesis (and the
onlyform,perhaps,thatWhorfwouldtodayrecognize)is thateach language
bothembodiesand imposesupon theculturea particularworldview.Nature
is, in reality,
a kaleidoscopiccontinuum,but theunitswhichformthebasis
ofthegrammar ofeachlanguageservebothtoclassify intocorresponding
reality
unitsand to definethefundamental natureofthoseunits.Thus, in English
and other"StandardAverageEuropean"tongues,thebasic unitsofrealityare
objects(nouns),composedofsubstanceand form,and actions(verbs)- both
ofwhichexistin an objective,three-dimensional space (expressedbysuchlin-
guisticdevicesas locatives)and a "kineticone-dimensional uniformly and per-
petuallyflowingtime"(expressedbyformssuchas tense).(2) In theHopi lan-
guage,however, thingsand actionsarenotdistinguished; rather,theyareboth
Events,differentiated onlyaccordingtoduration.EventosaythataboutHopi
maybe misleading,forratherthansubstance,motion,space,and time,Hopi
grammar dividestheuniversebytwogreat"principles," "Manifested" (Objective)
and "Unmanifest" (Subjective)."Manifested" comprisesall thatis or has been
accessibletothesenses,while"Unmanifest" (Subjective)includes,as one group,
all thatwe call futureand all thatwe call mental,includingthatwhichis per-
ceivedas future-potential-mental in the"heart"ofmen,animals,plants,inani-
mateobjects,and theCosmos. The metaphysics implicitin thegrammarof
StandardAverageEuropeanmakesit sensibleto analyzesentences,and thus
reality,intoagents,actions,and theobjects,instruments, and resultsofactions;
butsuchconstructions, Whorfargues,aregrossdistortions whenused as units
ofanalysisforvariousAmericanIndian languages.In supportofhis conten-
tions,Whorfprovidesa varietyoftranslations ofstatements in variousIndian

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
256 Et cetera • FALL1987

languagesintoEnglishto showhow unlikeours arethethoughtprocessesof


speakersof thoselanguages.Thus, in Apache, "It is a drippingspring"is
expressedby"Aswater,orsprings,whitenessmovesdownward." In Shawnee,
"cleaninggunwitha ramrod"is "directhollowmovingdryspotbymovement
of tool."
The Whorfianhypothesisis, at the least,intriguing:whatis it liketo live
in a mentalworldinwhichthereareno thingsor actionsbutonlyevents,where
thereare no agentsand acts,no separatespace and time?Can we everhope
to communicatewithpeople who havesuch a worldview?At themost,the
Whorfianviewchallengesourmostfundamental beliefs.Arecommon-sense
distinctions (such as that between objectand action)whichappeartous tobe
"given"unequivocallyby our senses,actuallyan illusionfostered bythegram-
marofEnglish?Is Newtonianphysicsnota necessaryfirststepin thedevelop-
mentofphysicaltheory,but merelya metaphorderivedfromthegrammati-
cal unitsofStandardAverageEuropean?Arethebasicconceptsoflinguistics
perhapstheonlymeansbywhichwe mayhope to surmountthelimitations
ofourownlanguageandbecomeabletoanalyzedifferences betweenlanguages,
or aretheseconceptsthemselvesonlyreifications ofthegrammarofourown
languagefamily?
Upon whatevidencearesuchsweepingclaimsbased?As a linguist,Whorf
foundthegrammarofseveralAmericanIndian languagesto differ fromEn-
glishgrammar to such an extent thatliteral
translationsbetween those languages
and Englishmade no sense. The literaltranslations, givenabove,of"a drip-
pingspring"and "cleaninga gunwitha ramrod"do,indeed,appeartobe prod-
uctsofa veryalienmodeofthought.Of course,itis also truethatall languages
have somewhatdifferent grammars,eventhe languageswhich Whorfcalls
"StandardAverageEuropean."However,noticethatwhen we learnFrench,
we aretaughttotranslate "Commentallezvous?"notliterally as "How go you?"
butas thestandardEnglishgreetingtowhichitcorresponds, "How areyou?"
And ifa studenttranslates "le chatgris"as "thecatgray"he is toldhe has made
an error;in English,modifierscome beforethe noun,notafter,and thecor-
rectrenditionofthe phrasein Englishis "thegraycat."Let us, however,try
totakea Whorfian viewofFrenchand supposethattheorderofnounandmodi-
fieris indicativeofa difference in metaphysics. The Frenchlanguage,we may
assert, defines the basic units of nature not as substantive thingsbutas perva-
siveattributes suchas colors,shapes,and sizes. Whatwe see as a thing-with-
attributes,theFrenchman seesas a specificlocalperturbation ofa generalattrib-
ute;thus, what we call a cat with a particular coloris, French,a particular
in
modification ofthegeneralcolormanifold-some "catgray"as opposed,per-
haps, to some "foggray."Why should such interpretations seem absurdfor
Frenchbut not Hopi? Is it thatwe have otherevidenceforconcludingthat
Frenchmen arenotthatdifferent fromourselves? Iftherewerea sovereign Hopi
nationto the southofthe United States,mightwe todaybe learningin our
classroomsnotto makeerrorsof literaltranslationin Hopi?

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
Linguistic Relativity 257

The argumentforlanguageas implicitmetaphysics is circular.It is circu-


lar because the onlyevidencethatwe haveforthe metaphysicsis thegram-
mar,yetwe are clearlyunwillingto interpret everygrammaticaldifference
betweenlanguagesas metaphysical. The circularity appliesjustas welltothe
weakerformoflinguisticrelativity as to thestrongerdeterministic hypothe-
sis.The seriousness ofthecircularitybecomesapparentwhenwe considerhow
it mightbe eliminated.Can we go out and gatherindependentevidencefor
thought differences? Whatshouldwe lookfor?The problemappearstobe con-
ceptualratherthanempirical.What would it "look like"ifa languagecom-
munityhad a particularimplicitmetaphysics?We can recognizeexplicit
metaphysics and differencesbetweenthem;forexample,Aristotle's metaphysics
wasdifferent fromHegel's,and theUpanishadscontaina different metaphysics
fromtheNew Testament.IftheHopi made explicitmetaphysical statements
("theCosmosconsistsoffourmutuallyantagonistic Substances," etc.),we could
contrast thosestatements withtheexplicitstatementsoflikekindmadebyother
communities. But an implicitmetaphysics? Afterall,theHopi actin thephys-
ical worldmuch likeanyoneelse; theywalkthroughspace, bump intosolid
objects,and keeptrackoftimein plantingand harvesting. How arewetoknow
thattheyconceiveoftheseabstractions in a manneruniquelycorresponding
to and/ordeterminedby theirlanguage?
In summary: themostdramaticformoftheWhorfianhypothesis -the asser-
tionthateach languageembodiesand imposesupon theculturean implicit
metaphysics -does not,in thatform,appearto be an empiricalstatement. If
itmustbe interpreted as meaningonlythatlanguagesdiffer, thenitis truebut
trivial.If it is to meanmorethanthat,we findthatwe haveno idea whatthe
stateoftheworldwould"looklike"ifthehypothesis weretrue,or,correspond-
ingly,ifitwerefalse.The restofthischapterdiscussessuccessiveattemptsto
reinterpret theWhorfianviewintoclaimswhicharesufficiently specificthat
we can understandtheirmeaningand testwhethertheyare trueor false.
Grammatical Form Class
The words(actually,themorphemes,or unitsofmeaning)ofanylanguage
canbe dividedintoclassesofgrammatical equivalentson thebasisoftheposi-
tionswhichtheycan occupyin wordsequences(suchas sentences).The most
basic units of grammar,which Whorf claimed formedthe basis of the
metaphysics ofa language,arenone otherthanthemostgeneralformclasses
oflanguage-in Englishthepartsofspeechsuch as nouns,verbs,adjectives,
and adverbs.Manyformclassesaremorelimitedin scopethanthebasic"parts
ofspeech":genderdefinesclassesofnounsin French;Englishnounsareeither
"mass"(occurin theposition"Some X") or "count"(occurin theposition"An
X"); and in Navajo,verbsofhandlingtakea different
formdependingon the
natureoftheobjectshandled.Obviously,formclassesarenotthesame in all
languages.
As longas formclassesareconsideredonly"structural"
(definedonlybyposi-

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
258 Et cetera • FALL1987

tionofoccurrence in sentences), theydo notsuggestimportant cognitive differ-


encesbetweenspeakersofdifferent languages.However, Whorf and others have
stressed thatformclassesalso havesemantic(meaning)correlates. Thus, nouns
areseenas substances;verbsas actions;massnounsas indefinite, uncontained,
flowing massesofmatter; countnounsas singular, self-contained objects;gender
as masculine,feminine, and neuter;and Navajo verb-stem classifiers
as shape
types(round,long,granular, etc.)Generally, the members ofa linguisticcom-
munity are unconscious ofthe semantics of form class. For example,evenin
a relatively grammatically self-conscious society like ours, most people have
neverspontaneouslynoticedthedistinctionbetweenmassand countnouns,
noreverthoughtaboutwhichEnglishverbscan orcannottaketheprefix "un-."
Whorfspeaksofthesemanticcorrelates offormclasses(he calls them"cryp-
totypes")as the"covertcategories," the"underlying concepts"ofthelanguage.
In fact,it is thepervasive,covertinfluenceofcryptotypes on thoughtwhich
may be one relativelyconcrete interpretation ofwhat it mightmeanforgram-
marto influencemetaphysics.
The semanticinterpretation of formclass has not gone unchallenged.
Descriptivelinguistics considers the relationbetweenstructurally definedform
classesand theirsemanticcorrelates highly dubious. (3) Semantic definitions
offormclass arealwaysunclearor overextended; notall nounsaresubstances
(e.g., "space") norall verbsactive(e.g., "hold"); mass nounscan come in dis-
creteunits("somebread"),and countnounscan referto fluidmasses("a mar-
tini");masculineand femininegenderformsare used forinnumerablegen-
derlessobjects;and specificNavajo shape classifiersareused forabstractions
("news"takesthe roundclassifier).
There is, however,undoubtedlya partial correlationbetweensome form
classesand somesemantics.It wouldbe totheadvantage ofindividuals learning
a languagetobe aware(at somelevel)ofthesepartialcorrelations. RogerBrown
(1958) has shownthateven4-year-old childrencan use structural syntactic cues
forguessingthesemanticreferent offormclasses. (4) Brownshowedthechil-
drenpicturesin whichan action,a discreteobject,and an unboundedflow-
ingmassweredepicted,introducing thepictureeitherwith"This is a picture
oflatting"or "ofa latt"or "ofsome latt."The 4-year-olds easilyidentified the
objectby means of the form-class cue. A similar was
experiment performed
on theformclass genderby Ervin. (5) Italianspeakerslivingin Bostonwere
readnonsensesyllablesformedwithItaliangender.Whensubjectswereasked
to ratethe syllableson a seriesof adjectivescales,theyratedthe masculine
gendersyllablesmoresimilartotheirratingsfor"man"than"woman"and vice
versa.(6) Such experiments demonstrate thatwe can makeuse ofwhatsemantic
information thereis in formclasseswhenwe arelearningand applyingwords.
They do not,however,provethatspeakersoflanguageswithdifferent setsof
formclassestakedifferent viewsofthesemanticnatureoftheworld.Afterall,
discretesolid objectsand unboundedfluids,and male and femaleorganisms,
havequitedifferent physicalproperties whichall peoplesmightwellbe required

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
Linguistic Relativity 259

totakeequal accountofwhetheror nottheirgrammar makessuchdistinctions.


Ifwe wishedtotestwhethersemanticaspectsofformclassdo affect thought,
whatkindsofcorrelates or effectson thoughtmightwe lookfor?In fact,there
hasbeenlittlesystematic consideration and littleresearch concerningthisissue.
One possibilityis thatthereis a "metaphoricalgeneralization" ofthe mean-
ing frommembersofthe formclass to whichit literallyapplies to members
to whichit does notapplyliterally at all. Thus, theFrenchmayreallythink
ofand treattablesas feminine, and theNavajo mayconsidernewstobe round.
Whorfhimselfsuggeststhiskindofinterpretation whenhe claimsthatwe read
actionintoall wordsthatare verbs,and, since all Englishsentencescontain
verbs,intoeverystatement. "Wetherefore readactionintoeverysentence,even
into'I hold it' . . . We thinkof it (i.e., holding)and even see it as an action
because languageformulates itin thesame wayas itformulates morenumer-
ous expressions, like'I strikeit,'whichdealswithmovements and changes."(7)
But do we readactionintoall verbs?How can we tell?One téstwould be to
go tothenaturallogicoflanguageuse itself;ifactionis being"readinto"verbs
like"hold,"theyshouldbe capable ofoccurringmodifiedby actionadverbs
justas do "true"actionverbs.The actual stateof such verbsis describedby
thephilosopherMax Black: "a man maystrikeslowly,jerkily,energetically,
and so on. Now ifsomebodywereto attachtheseadverbsto theverb'to hold,'
thatwouldbe sufficient indicationthathe was 'readingaction'intotheverb.
I supposea childmightsayhewasholdinghishatslowly, andthepoetis allowed
a similarlicense; but otherwisethe conceptual confusionis too grossto
occur."(8)
Arethereanycasesin whichthepartially correlated
semanticsofa formclass
areextendedtootherwordsthathappentobe in thatclass?Is therea systematic
wayofstudyingsuchextensionsso thatwe mightconcludethatitneverhap-
pens?These intriguing questionsremainentirely open tofutureinvestigation,
and theinterested studentmightwell tryusing his intuitionas a speakerof
his own languageto considerthem.
Even ifthe semanticpartialcorrelatesof formclass do notextendbeyond
theclear-cutcases,theymayhaveeffectson thought-one obviouspossibil-
ityis thattheycontinuallydrawthe attentionof speakersofthelanguageto
thoseaspectsoftheworldwhicharethebasis forthe(evenpartial)form-class
semanticdistinctions.Such an effectwould be mostlikelydemonstrablein
thecase ofa formclass whichwas sufficiently salientlinguistically
and suffi-
cientlycorrelatedwitha clear-cutsemanticforit to be reasonableto expect
speakersofthelanguagetobe influencedbyhabitualuse oftheclass. Navajo
shape classifiersappearto be justsuch a case. Use ofverbstemswhichindi-
cateshapein Navajo is obligatory linguistically,and suchstemsareveryhigh-
frequency itemsin thelanguage.For all objectsthatactuallyhavea shape,the
classifiersareusedconsistently- evenEnglishspeakersnewtothesystemper-
ceivetheclassifiers to referpredictably to shape types.Most importantly, the
classifierused foran objectis not an invariableattributeof the object(as is

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
260 Et cetera • Fall 1987

generallythecase withgender)but varieswiththeactualshape oftheobject


at thetimeof reference. Thus, a rolled-uprugwill takea different classifier
fromthesame rugwhenitis lyingflat(S. Ervin-Tripp, personal communica-
tion, 1972). It is reasonableto suppose thatNavajo speakersarecontinually
noticingshapewhenspeakingand,thus,wouldbe morelikelytonoticeshape
and touse shape as a basis ofclassification thanspeakersofa languagewhich
does not incorporateobligatorygrammaticalshape distinctions.
A testforthishypothesis was devisedbyCarrolland Casagrande(1958). (9)
Subjects were presented with a reference objectand twootherobjects,each
one resemblingthe standardby a different attribute.For example,the stan-
dardmightbe a redcircleand theotherobjectsa redsquareand a blue circle.
Subjectswereaskedtochoosewhichofthetwoobjectswas mostlikethestan-
dard. A varietyof objectswereused, incorporating all combinationsofthe
attribute's form,color, and size. The basic hypothesis thatNavajospeakers
was
wouldprefer toclassifybyformratherthanbytheotherequallycorrectattrib-
utes.Three groupsofchildrenweretested:Navajo childrenwhosedominant
languagewas Navajo; Navajo childrenforwhom Englishwas thedominant
language; and monolingualEnglish speakers.As predicted,the Navajo-
dominantNavajo preferred to classifyon the basis of form;however,so did
theEnglish-speaking Bostonchildren.English-dominant Navajo,on theother
hand,preferred color.These results arenot unequivocal.Theyhavebeentreated
bothas evidenceagainsttheWhorfianview- becauseEnglishspeakers,whose
grammardoes notcall attention to form,preferred formclassifications justas
muchas theNavajo-dominant Navajo -and as evidence in support ofWhorf-
becausetheEnglish-dominant Navajo,whosecultureand earlyenvironment,
butnotlanguage,werethemostsimilartothatoftheNavajo-dominant Navajo,
preferred colorand notform.The readermightpauseand considerwhatargu-
mentsmightbe offeredon each side.
In fea, theresultsareprobablyevenmoredifficult tointerpret thanhasbeen
supposed. Since thetime of Carroll and Casagrande'sexperiments, a greatmany
testsofcolor-form preference in classification have been performed on a vari-
of
ety populations, and a consistent but, to the presentauthor,incomprehen-
sible patternofresultshas emerged.Aroundtheworld,theyoungerthesub-
ject and/orthe less Westernschoolinghe has received,the morelikelyhe is
touse coloras thebasisofclassification in thekindof"triads"testused byCar-
rolland Casagrande.(10) However,itis justyoungchildrenand non-Western
peoples who appear to be the populationsleast likelyto classifyby color in
morenaturalistic contexts. It is thetechnologically less-advanced cultures which
appear to have smaller color vocabularies and less cultural concern with color
distinctions and coordinations.(11) Youngchildrenin Westernculturesonly
cometouse colortermscorrectly and consistently at abouttheage whenthey
to
begin prefer form classification in the color-form preferencetriads.(12X13)
Furtherevidencecomesfroma studyofthepublisheddiariesofthelanguage
developmentof individualchildren(usuallykeptby fondlinguistparents).

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
Linguistic Relativity 261

Clarkhas examinedall ofthediaryexamplesofyoungchildren'sovergener-


alizationsofwords-that is,ofcaseswherea childappliesa newlylearnedword
to a varietyofthingsto whichthatworddoes notactuallyapplyin theadult
language(e.g., callingall animals"dogs"or all men"daddy").(14) Clarkwas
interested in findingout theattributes bywhichchildrengeneralizemeaning.
Whatis relevanttothepresentissue is that,in all ofthediaryliterature, there
is notone singleinstancereportedin whicha child seemedto overgeneralize
a wordon thebasis ofcolor!So classifying bycoloror formon thetriadtype
oftestmaywell be theresultof factors veryspecificto operationsand cogni-
tionsin thattestsituationand maynotreflect anytendencytouse eitherattrib-
ute as the basis of classificationin anyothercontext.
Navajo shape-classifying verbstemsappearedto be a case ideallysuitedto
a demonstration of the effectof grammaticalformclass on attention.How-
ever,preference forformovercolor in a triad-classification taskmayinvolve
toocrudea hypothesis ormaybe tootask-specific totesttheissue.The interested
readermighttrytothinkofa morereasonabletest.Does anythingin theliter-
atureon perception, memory, learning,problemsolving,or otherhumancog-
nitivefunctionssuggestsuch a test?Unfortunately, at present,the evidence
concerningthe effectsof formclass on attentionremainsequivocal.
If the semanticsof formclass providethe cryptotypes -the underlying
categorization ofreality-forspeakersofa language,thenat theveryleast,at
somecognitivelevel,speakersofthelanguageshouldcode formclassesas cat-
egories.Overtsemanticcategorieshave been shownto have severalreliable
effects on humanmemory-do categories consistingofformclasseshavesimilar
effects?
One sucheffect occursin thekindofexperiment called "freerecall."A sub-
ject is read a listof ordinarywordsand thenattemptsto recallthewordsin
anyorderhe wishes.The listmayconsistof random,unrelatedwords,or it
maycontaina numberofwordsfromthesame semanticcategories(forexam-
ple,flowers, animals,musicalinstruments). When subjectsreceivea "catego-
rized"wordlist,theyremembermorewordsthanin uncategorizedlists,and
theytendtorecallthosewordsin "clusters" fromthesamecategory-evenwhen
theinputlistcontainèdthewordsin randomorder.If grammatical classesare
"meaningful" cognitive categories,shouldn'tthey also provide a basisforcluster-
ing and improved recall? Cofer and Bruce (1965) presented listsin which words
could be categorizedintothe formclasses nouns,verbs,and adjectives,but
wereotherwise unrelated.(15) They foundno effects ofthecategories on either
or
accuracy clustering in recall.(16)
A secondeffect ofsemanticcategorieson memoryarisesin thetypeoftask
in whichsubjectsmusttryto rememberan item(such as a number,word,or
nonsensesyllable)forshortperiodsoftimewhileperforming an interfering
task, such as counting backwards. In such experiments, abilityto remember
becomesrapidlypoorerwitheachsucceedingitem,generally attributed tointer-
ference frompreceding items.Ifall oftheitemsup tosomepointhavebelonged

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
262 Et cetera • FALL1987

toone semanticclassand theexperimenter switchessuddenlytoanotherclass


(e.g.,switches from letters
ofthealphabet numbers,fromanimalstoplants,
to
or fromwordswitha "good"connotationtowordswitha "bad" connotation),
subjectsregaintheirabilitytoremembertheitems.However,switchingfrom
one grammaticalclass to another(e.g., fromverbsto adjectives)has no such
effects.(17) So, at leastin thesememorytasks,grammatical classseemstobe
morelikea dead metaphorthanlike a psychologically real classification of
reality.(18)
We beganthissectionwiththenotionthatthesemanticcorrelates ofgram-
maticalformclass mightprovidethosebasic classifications ofrealitywhich
arethecovertmetaphysics embodiedand perpetuated bylanguage.However,
we havebeen unableto verify thespecificmeaningswhichsucha claimmight
have.Membersofgrammatical classeswhichdo notsharethe"semanticmean-
ingoftheclass" also do notseemtobe treatedmetaphorically as thoughthey
did shareit by otheruses in thelanguageitself.The one attemptto testthe
hypothesisthatspeakersof a languageare led to pay particularattentionto
attributesoftheenvironment coded in obligatoryformclassesin languagewas
rendereddifficult to interpret a
by variety of factors.
Finally,thereis someevi-
dencethatformclassesdo notseemtobe codedas cognitive categoriesat all- at
least,they do not have the that
properties other meaningful semantic categories
do. However,it shouldbe apparentfromtheprecedingsectionsthattheevi-
dence now availableis anythingbut systematic.
Most actual researchdirectedtowardthe Whorfianhypothesishas not
addresseditselfto grammarat all, but has dealt withthe relatively concrete
and overtclassifications madebylanguagevocabulary. It is tothattopicwhich
we turnnext.
OVERT LANGUAGE CLASSIFICATIONS: VOCABULARY
According toWhorf,languageaffects thoughtbasicallybymeansofthekinds
ofclassifications it "laysupon" reality.Whorffocusedon classificationsofa
generaland abstract nature-thecovert"metaphysics" and"cryptotypes"
embed-
ded in language.However,we haveseen thedifficulty ofdemonstratingthat
classifications on thatlevelareactuallyrelatedto meaningfulcognitiveunits.
Thereis anotherleveloflanguage,however, in whichsemanticclassifications
are quite overt,thelevelofthelexicon(vocabulary).If the same Hopi word
refersto whatEnglish codes withthe threewords"airplane,""insect,""avia-
tor,"or ifthe languageof theEskimo uses threewordsto referto thatthing
coded by the one Englishword"snow,"theseare overtsemanticdifferences
in thewaythe worldis "cutup" and coded.
In fact,it is notunreasonableto suppose thatthereare concreteand iden-
tifiableaspeas ofthelexicalcode thataffectidentifiable
and measurableaspects
of thought.Earlier in this book, the backgroundfora clear-cutempirical
demonstration ofjustsuch a case was presented.Berlynedefinedtheconcept
of"information" and discussedhumanlimitationsin thecapacityto process

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
Linguistic Relativity 263

and retaininformation. How muchinformation can be retained,however,is


dependenton thewayin whichit has been "coded."Normally,forexample,
onlyaboutsevendigitscan be remembered, evenfora verybriefperiod.(19)
However,ifa stringof0 and 1 digitsis recodedbya subjectintooctal(a system
bywhichgroupsofthreedigitsare"named"bya singlenumber),he can remem-
ber almostthreetimesas many.Here is a prototypical case in whicha clas-
aspectofvocabulary
sificatory (thenumberofdigitsnamedbya singlenumber
in a particularcode) can be showndirectlyrelatedto an aspect ofcognition
(amountof information storedin memory).
Octalis an "artificial"code developedforparticularpurposeswithinourcul-
ture.The Whorfianhypothesison the levelof the lexiconmightbe consid-
eredtocontainan additionalassertion,namelythatnaturallanguagesas they
arespokenbytheworld'sspeakerscontaindifferences in codes(analogous,per-
haps,to recodinggroupsofdigitsintooctal)whichaffectcognitiveprocesses
such as perception,classification, and memory.
The bulkoftheempiricalworkon linguisticrelativity has involvedlanguage
at thelexicallevel,and the restofthischapterwill be concernedwithwork
at thispointto pause and considerhowsuch
at thislevel.It is well,therefore,
a hypothesis aboutnaturallanguagescan (and cannot)be tested.In theprevi-
ous sections,we saw in operationvariousproblemsconcernedwithtesting
Whorfian views;atthispoint,letus looksystematically attherelevant methodo-
logicalissues.
A Discussion of Method
Manyfactswhichhavebeen offered in supportoftheeffects oflanguageon
thought(at all levelsof language)havebeen onlydescriptionsofdifferences
betweencultures.To avoid such confusions,it is necessaryto bear in mind
theimportant distinction betweenthecontent ofa languageor cultureand the
thought procedí ofmembersoftheculture.Of course,culturesdiffer in con-
tent;we wouldprobablynotcall themdifferent cultures theydid not.A rice
if
farmerin the Philippinesand a collegestudentin Americalive quite differ-
entlives,and presumably thecontentoftheirthoughts, knowledge, andmemory
mirror thosedifferencesin experience.But fromknowingthat,we cannotauto-
maticallyassumethatmembersofthetwoculturesoperateon thatcontentin
differentways.It is probable,forexample,thattheyforgettheirexperiences
accordingto thesame laws ofdecayor interference in memoryregardlessof
whatit is thattheyare forgetting.
Thereis,ofcourse,a sensein whichanylexicaldifference betweenlanguages
impliesa difference in thecontentofthoughtofthespeakers.In learningto
use a term,speakersmustlearntheclass ofthingsto whichthetermrefers;
thus,knowledgeofand reference to thatclass ofobjectsis partofthecontent
ofthespeaker'sthought.In thissense,theweakerformoflinguisticrelativity
(thatthereare differences in thoughtin different linguisticcommunities)is
necessarilytrue.The reallyinteresting hypothesisat thelexicallevel,however,

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
264 Et cetera • FALL1987

is the strongerdeterministic claim thatlexicaldifferences themselvesaffect


thoughtprocesses in some manner.
It is tempting,when makingclaims supportingthe Whorfianhypothesis
at anyleveloflanguage,torelyprimarily on contentdifferences.Theyareoften
verystrikingdifferences; ifa languagehas onlytwocolortermsor thousands
ofelaboratedistinctions and classifications forskindiseases,surelythatmust
affect thewayin whichthesedomainsaredealtwithbythecognitivemanipu-
lationsofthe speakers.To illustratehow misleadinga directinference from
lexicalcontentcanbe,we mayrecalltheevenmorestriking differencesbetween
the Hopi grammaticalclassifications of thingson thebasis ofdurationand
theEnglishdivisionintosubstancesand actions.The evidenceconcerning that
distinction leftus in gravedoubtaboutwhethernounsandverbsaremeaningful
semanticcognitivecategoriesforEnglish speakersat all.
Most "demonstrations" ofthe Whorfianhypothesishave done morethan
simplypoint to differences between thecontentoflanguages;theyhave,inaddi-
tion, identifiedaspects of the culture ofthespeakerswhichcovarywithlan-
guage. Such evidence is not entirely adequateeither,however, fortworeasons.
In thefirstplace,covariation does notdetermine thedirectionofcausality.On
thesimplestlevel,culturesare verylikelyto havenamesforphysicalobjects
whichexistin theircultureand notto havenamesforobjectsoutsideoftheir
experience.Wheretelevisionsetsexist,therearewordsto refertothem.How-
ever,it would be difficult to arguethattheobjectsare caused by thewords.
The samereasoningprobablyholdsinthecase ofinstitutions and other,more
abstract, entitiesand their names. In the second place, covariation between
culturalcontentand languagecontentneitherprovesthefurther existenceof
covarying cognitiveprocessesnorwould itdeterminethedirectionofcausal-
ity, even were such covariationto be demonstrated. Thus, if Eskimoswere
shownbothto havemorenamesforsnowthanAmericansand to remember
different typesofsnowbetterthanAmericans,bothmightsimplybe due to
thefeetthatthereis moresnowin theArcticand toEskimoshavingmoreactive
experiencewithit thanAmericans;it would not havebeen provedthatthe
greaternumberofwordsper se affected the memory.
The precedingargument has stressedthepointthatcognitive processesmust
be measuredindependently of, and not simply deduced from,linguistic orcul-
turalcontent.However,thisraisesa secondmajorproblemofmethod:how
are we to defineand measurecognitiveprocessescross-culturally? Too often
such measurementis based on a psychometric "deficit"model. Hypotheses
arestatedin termsof"howwell"entireculturesperform on a particulartest.
For example,a hypothesismightstatethat"membersoftraditionalcultures
cannotthinkcreatively" orthat"themorewordsa languagehas forcolors,the
betterspeakerscan remember colors."The investigatormightadminister a test
of "creativethinking"to Americans(not a traditionalculture)and to the
Yemenites(a traditionalculture),or mightadministera colormemorytestto
Americans(manycolorterms)and to the Dani (fewcolorterms).When the

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
Linguistic Relativity 265

Yemenitesperformed poorlyon thecreativity testand theDani poorlyon the


colormemorytest,theinvestigator would concludethathis causal hypothe-
siswas supported.However,itshouldbe obviousthatinnumerableotherfee-
torsbesidesthosein whichtheinvestigator is explicitlyinterested varybetween
"us"and "them."Motivations, culturalmeaningfulness ofthematerials, general
familiarity with,or evenpreviousexplicittrainingwith,thekindoftaskused
aresomeobviousexamples.In feet,anypreliterate culturewill probablyper-
form"lesswell" thana WesternculturegivenalmostanyWestern"test."But
ifDani can be expectedtoperform belowAmericansin anymemorytest,how
maywe concludethatitwasthenumberofcolortermswhichdetermined their
poorperformance inthecolormemorytest?In short,positiveresults areassured
theinvestigator who frameshypothesessuchthata singleWesternand single
non-Western cultureare compared,witha predictionin the directionofthe
non-Western culturegivingpoorerperformance thantheWestern-but such
resultswill be uninterpretable.
Aretherewaysoutoftheimpasse?One trendhas been to tryto inventtasks
whichareas culturally relevantin contentand formofadministration toa par-
ticularpreliterate cultureas Westerntestsareto Westerncultures.This excel-
lentidea has, however,givenriseto a special sortofcircular"dialectic."The
formatoftheresearchis typicallythis:Stage 1-an investigator demonstrates
thatthepeopleof"CultureX" feiltoexhibitsomeability(forexample,"abstract
thinking") on a standardWesterntest.Stage2 -the same or a different inves-
tigator manipulates thecontentand context ofthetestuntilhe hasdemonstrated
that,undertherightcircumstances (forexample,ifaskedtoreasonaboutanimal
husbandryin theirown cultureratherthanabout coloredgeometricforms),
thepeople of CultureX do exhibit"abstractthinking." The Stage2 demon-
stration maybe beautifulin itsingenuity;however, thetwostagestendsimply
tocanceleach otherand makelittlecontribution toourunderstanding ofbasic
humanthoughtprocesses.It oughttogo withoutsayingthatall tasksin cross-
culturalresearchshouldbe as appropriateforthepeople performing themas
possibleand,indeed,somelevelofappropriateness is essentialifanymeaningful
dataareto be collectedat all. However,culturallymeaningfultasksdo notof
themselves producewell-conceived whyshouldhypotheses
research; be framed
intermsofdifferences in absolutelevelofperformance between"us"and"them"
at all?
A secondapproachto theproblemuses a modelbased on epidemiological
methodsin thehealthsciences.If traditionalism wereproposedto be related
tocreativethought,manydifferent cultureswhichdiffered in degreeoftradi-
tionalismmightbe studied,as mightgroupswithinthesame culturewhich
differed in degreeoftraditionalism. Culturesand groupstobe comparedwould
ideallybe chosen so as to hold constantvariousotherfactorswhich might
influencecreative thought.(20X21)Unfortunately, owingtopracticalconsider-
ations,theepidemiologicalmethodhas seldombeen applied to thestudyof
cognitivefactors.

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
266 Et cetera • FALL1987

Perhapsthesimplestand mostdirectwayofcircumventing theproblemof


measuringcognitive variables cross-culturally is to abandon research designs
whose emphasisis on "main effects"of cultureper se. Hypothesescan be
formed,notin termsofabsolutedifferences betweencultures,butin termsof
interactions between variables within and between cultures.Take,forexam-
the
ple, hypothesis that the number of colorterms affectscolormemory. Instead
ofcomparingspeakersoftwolanguages,one ofwhichhas morecolorterms
thantheother,we mightsearchforcases whereitis possibleto comparerela-
tiveperformance fordifferent areas of the color space forlanguageswhich
differed in therelativenumberoftermstheyhad fortheseareas.Perhapsone
languagehas manytermsforblueand greencolorsbutfewtermsfortheyellow-
browncolor area,anotherlanguagejustthe opposite.Our predictioncould
thenbe thatspeakersofthefirst languagewouldshowrelatively bettermemory
forthe blue-green than forthe yellow-brown area, whereas speakersof the
secondlanguagewouldbe relatively moreproficient withyellow-brown colors
thanwithblue-green. Withresearchso designed,itwouldnotmatterhowwell
eitherculturerememberedcolortermsin total.Such an approachmaybe a
keyto meaningfulcomparisons,evenbetweenquite different cultures.
To returnto the Whorfianhypothesis:it should by now be apparentthat
manyfactorsarenecessaryin orderto havea realtestoftheeffect ofa natural
languages lexicon on thought,(a) We must have at leasttwo natural languages
whoselexiconsdiffer withrespectto somedomainofdiscourse- iflanguages
are notdifferent, thereis no pointin the investigation. (22) (b) The domain
must be one which can be measured by the investigator independently ofthe
way it is encoded by the languages ofconcern (forexample, color may meas-
be
uredin independentphysicalunitssuchas wavelength)-if thatis notthecase
(as, forexample,in such domainsas feelingsor values),thereis no objective
wayofdescribinghowitis thatthetwolanguagesdiffer, (c) The domainmust
notitselfdiffer grossly between the cultureswhose languagesdiffer-ifitdoes,
thenit maybe differences in experiencewiththedomain,and notlanguage,
whichareaffecting thought,(d) We mustbe able toobtainmeasuresofspecific
aspects cognition-such as perception,memory,or classification
of -having
to do withthedomainwhichareindependentof,ratherthansimplyassumed
from,thelanguage,(e) We musthave a cross-culturally meaningfulmeasure
ofdifferences in theselectedaspeas ofcognition-preferably we shouldbe able
to statethe hypothesesin termsof an interaction betweenthelinguisticand
cognitivevariables,ratherthanin termsofoveralldifferences betweenspeakers
ofthe languages.
One domainonlyhas appearedto researchers tobe ideal forsuchresearch -
color.Color is a continuousphysicalvariablewhichcan easilybe designated
byobjectivemeasureswhichareindependentofthecolortermsin anygiven
language.Many reportsin theanthropological literature havedescribeddiffer-
encesin colorterminologies betweenlanguages- thatis,differences in theway
in whichcolor termsappear to classifythe physicallyinvariantcolor space.

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
Linguistic Relativity 267

The physicalaspectsofcolor,thedomainofcolorsas such,is thesamein every


partoftheworld- although,ofcourse,thecolorsmostfrequently viewedmay
differ ecologically.Color discrimination,memory,and classification can be
readilymeasuredindependently ofcolornames,ratherthansimplyinferred
fromthecolorterminology colorslendthemselves
oftheculture.And,finally,
readilytohypotheses - as has alreadybeen illus-
statedin termsofinteractions
tratedby the precedingexamples.
It mayseem a longwayfromtheinitialintroduction oflinguisticrelativity
as an assertionaboutdifferences in "worldview"toa studyofthepossiblecog-
nitiveeffectsofdifferences in colorterms.The transitionwas madenecessary
bytherequirement thatassertionsbe madein theformofempirically testable
hypotheses.Much of the remainderof thischapterwill tracethe historyof
language-cognition researchin thedomainofcolors,theprimarydomainin
whichsuch researchhas been carriedout.
Color
Color is perceivedwhenthehumanvisualsysteminterprets certainaspects
ofthephysicalpropertiesoflight.Sensorypsychologists describecolorwith
a solidusingthreepyschological dimensions:hue (roughly, thedominantwave-
lengthofthelight),brightness (looselyspeaking,an intensity dimension),and
saturation (theapparentdegreeofdominanceofthedominantwavelength, the
"purity"of the light).(23) The color solid is divisibleintoliterallymillions
ofperceptuallyjustnoticeabledifferences. There is no evidencethathuman
populationsdifferin the physiologyofthevisual system,northatthereare
anycultureswhichdifferin actual abilityoftheirmembersto perceiveand
discriminate colors.(24) In feet,thereis evidencethattheold worldprimates,
whosecolor physiologyis similarto thatof humans,are not different from
humansin colorperceptionand discrimination. (25) There arefarfewercolor
namesin anylanguagethantherearediscriminable colors,and fewerstillcom-
monlyused colornames.Thus, it appearsthatculturaldifferences are to be
foundon the levelof categorization ratherthanperceptionof color.
A seminalstudyin the effects of languageon cognitionwas performed by
Brownand Lenneberģ.(26) Brownand Lennebergreasonedthatcultures,per-
haps because of differing color "ecologies,"should differwithrespectto the
areasof the color space to which theypaid the mostattention."Culturally
important" colorsshouldtendtobe referred tooftenin speechand,thus,their
namesshouldbecomehighly"available"tomembersoftheculture."Availabil-
ity"ofa nameshouldhavethreemeasurableattributes: as Zipfhasshown,words
used frequently tendto evolveinto shorterwords(forexample,automobile
becomesauto or car); thus,the lengthofcolor wordsshould be an indexof
theiravailability. (27) Secondly,a moreavailablewordshouldbe one which
a speakercan producerapidlywhenaskedtonamethethingtowhichtheword
refers.Finally,wordsfrequently used in communication shouldcometo have
meaningswidelyagreeduponbyspeakersofthelanguage.These threeindices

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
268 Et cetera • Fall 1987

ofavailabilityare linguistic measures;fora measureofcognition,Brownand


Lenneberg chose recognitionmemory,the abilityof subjectsto recognizea
previously viewed colorfromamongan arrayofcolors.The hypothesis relat-
the and
ing linguistic cognitive variableswas similar tothecase,describedprevi-
ously,oftheeffectiveness ofoctalas a code fordigits;nameswhichare more
availableshouldbe moreefficient codes forcolors(youcan hangon to them
betterin memory)-thus,people should be able to retainthemlonger.
Brownand Lenneberg'sactualexperiment was in twostages.Theyfirsthad
a sampleofEnglish-speaking Americanundergraduates namea sampleofcol-
oredchips.Becausethethreelinguistic measureswerefoundtocorrelate highly
(thesamechipstendedtobe givenshort,rapid,agreed-upon names),themeas-
ureswerecombinedintoa compositemeasurewhichBrownand Lenneberg
named"codability." Othersubjectsperformed a memorytask;theywereshown
eitherone or fourcolors,waiteda predetermined lengthoftimeduringwhich
thecolorswerenotvisible,and thenattempted to pickoutwhichcolor(s)they
had seenfroman arrayofmanycolors.The hypothesis was thatthemorecod-
able colorswouldalso be thebestremembered. That is exactlytheresultwhich
was obtained.Furthermore, theadvantageofthemorecodablecolorsincreased
as the numberof colorsand the lengthof timetheyhad to be remembered
increased.This studyis theclassicdemonstration ofan effectoflanguageon
memory.(28)
The Brownand Lennebergstudyused onlyspeakersofa singlelanguage.
However,itslogiccan easilybe extendedtoa cross-cultural comparison.Which
particular colors aremost codablewould be expected varybetweenlanguages,
to
but the lawfulrelationshipbetweenmemoryand codabilityshould remain
true- thosecolorswhichare mostcodable should be betterremembered by
speakersofthatlanguagethanthelesscodable colors.That thiswouldbe the
caseseemedso obviouslytruethatitwasnottestedformanyyears.Is itobvious?
The first,almosttrivialrequirementfortestingtheWhorfianhypothesis
whichwe listedpreviously wasthattherebe atleasttwonaturallanguageswhose
terminologies with respecttosome domainweredifferent. The anthropolog-
ical literature containsmanyreportsof such differences in color names- for
example,cultureswhichhaveonlyone wordto describethecolorswhichEn-
glishdistinguishes as "green"and "blue,"or cultureswhosewordfor"orange"
includesmuchofwhatwe wouldclassifyas "red."Fromthiskindofevidence,
it appearedthatlanguagescould arbitrarily cut up thecolorspace intoquite
different categories. two
Recently, anthropologists havechallengedthisassump-
tion.
Berlinand Kay firstlookedat thereported diversity ofcolornameslinguisti-
cally, and claimed that there were a
actually very limited numberofbasic- as
to
opposed secondary- colorterms in anylanguage.(29) "Basic" was defined
by a listof linguistic criteria:
forexample,thata termbe composedofonly
a singleunitofmeaning("red"as opposedto"darkred"),and thatitnameonly
color and not objects("purple"as opposed to 'Svine").Using thesecriteria,

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
Linguistic Relativity 269

Berlinand Kay reportedthatno languagecontainedmorethan11 basic color


names:threeachromatic(in English,"black,""white,"and "gray")and eight
chromatic (in English,"red,""yellow,"
"green,""blue,""pink,""orange," "brown,"
and "purple").
Berlinand Kay nextaskedspeakersof different languagesto identify the
colorstowhichthebasic colornamesin theirlanguagereferred. Their initial
groupofsubjectswere20 foreign studentswhosenativelanguagewas notEn-
glish.Subjectssaw a two-dimensional arrayofcoloredchips-all ofthehues
at all levelsofbrightness(all at maximumsaturation)availablein theMun-
sellBook ofColor. (30) The studentsperformed twotasks:(a) theytracedthe
boundariesofeach oftheirnativelanguage'sbasic color terms,and (b) they
pointedto thechip whichwas the bestexampleofeach basic term.As might
havebeen expectedfromtheanthropological literature, therewas a greatdeal
ofvariationin theplacementofboundariesoftheterms.Therewas not,how-
ever,reliable variation.
Speakersofthesamelanguagedisagreedwitheachother
in placementoftheboundariesas muchas did speakersofdifferent languages;
and thesameperson,whenaskedto map boundariesa secondtime,was likely
to map themquitedifferently fromthewayhe had at first.It is, thus,likely
thatevenanthropological reportsofdifferences in theboundariesofcolorterms
are confoundedby thisunreliability. Surprisingly, in spiteofthisvariation,
thechoiceofbestexamplesofthetermswas quite similarforthespeakersof
the20 different languages.Berlinand Kay calledthepointsin thecolorspace
wherechoicesof best examplesof basic termsclustered"focalpoints,"and
arguedthatthepreviousanthropological emphasison cross-cultural differences
in colornameswas derivedfromlookingat boundariesofcolornamesrather
thanat color-namefocalpoints.
Brownand Lenneberg'sresultshad been interpreted as a demonstration of
theeffectof codabilityon memory.However,Berlinand Kay's focalpoints
suggested a disturbingalternative.
Supposethatthereareareasofthecolorspace
whichareperceptually more"salient"to all peoplesand thattheseareasboth
becomemorecodableand can be betterremembered as thedirectresultoftheir
salience.The present authortestedthispossibility. (31) Ifcodabilityis theresult
ofsalience,thesamecolorsshouldbe themostcodable in all languages;spe-
cifically,focalcolorsshouldbe universally morecodablethannon-focal colors.
A focalcolorrepresenting each oftheeightbasic chromatictermswas chosen
fromthecenterofeach ofthebest-example clustersproducedby Berlinand
Kay'ssubjects;non-focal colorswerechosenfromthe"internominaP areasof
thecolorspace,areaswhichwereneverpickedas thebestexampleofanybasic
colorname. (32) The chipsweremountedon cardsand shown,one at a time,
in scrambledorder,individually, to 23 people whosenativelanguagewas not
English.A subject'staskwas to writedownwhathe would call each colorin
his language.The resultsofthestudywereclear:thefocalcolorsweregiven
shorternamesand namedmorerapidlythanwerethenon-focal colors.Thus,
in 23 diverselanguages,drawnfromsevenofthemajorlanguagefamiliesof

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
270 Et cetera • FALL1987

the world,it was the same colorsthatweremostcodable.


Therewas a secondparttothehypothesis. If memorywerethedirectresult
ofsalienceratherthanofcodability, focalcolorsshouldbe betterremembered
thannon-focal, evenbyspeakersofa languagein whichthesecolorswerenot
morecodable. Berlinand Kay'sclaim aboutthenumberofbasic colorterms
wasthattherewerenevermorethan,butcouldbe fewer than,11 terms;in fact,
theyargued that colorterms entered in a
languages specificevolutionary order.
The Dani of WestIrian (IndonesianNew Guinea) are a Stone-Age,agricul-
turalpeoplewho havea basicallytwo-term colorlanguage.(33X34X35)Color
systems ofthat character havebeen reported forotherculturesas welland form
StageI, the firstand simpleststage, ofBerlin and Kay's proposedevolution-
aryordering ofcolor For
systems. Dani, the eightchromaticfocalchipswere
notmorecodablethantheinternominal chips(establishedbyhaving40 Dani
nameall ofthecolorchipsin theBerlinand Kay array).WouldDani, neverthe-
less,betterremember thefocalcolors?To findout,Heider(1972c)administered,
toa sampleofDani anda sampleofAmericans, a colormemorytestverysimilar
toBrownand Lenneberg's.(36) Subjectswereshownfocaland non-focal colors,
individuallyin randomorder,for5 seconds,and aftera 30-secondwait,were
askedto recognizethecolortheyhad seen froman arrayofmanycolors.The
meannumbercorrectly recognizedbypeople ofeach cultureforeach kindof
chip is shown in Table 1. The mainresultswereclear:Dani, as wellas Ameri-
cans, recognized the focalcolorsbetterthanthe non-focal.
This studyalso illustratesa point about methodwhich was emphasized
earlier.A striking aspectofTable 1 is thatDani memoryperformance as a whole
was poorerthanAmerican.If the hypothesishad been in termsof absolute
differences betweencultures,we would havenotedthatDani bothhad fewer
colortermsand poorermemoryforcolorsthanAmericans,and mighthave
claimedthatlinguisticrelativity was therebysupported.However,itmustbe
remembered thattheDani area preliterate people,livingin face-to-face com-
munities,probably without need for or in
training techniques copingwith
for
the kind of overloadsof informationwhich this unfamiliarmemorytest
required.All ofthoseextraneousfactors undoubtedlyaffected Dani memory
performance as a whole. Our hypothesis,however, concerned differential
memory for different types of color within culture was not
and, therefore,
negatedby general culturaldifferences in "test taking."

TABLE 1
Accuracyof Color Memory:
Mean Number ofCorrectlyRecognized Colors
Stimulus colors
Culture Focal Internominal
U.S. 5.25 3.22
Dani .47
2^05

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
Linguistic Relativity 271

Color initiallyappearedtobe an ideal domainin whichto demonstrate the


effects oflexicaldifferences on thought;instead,itnowappearstobe a domain
particularly suitedto an examinationoftheinfluenceofunderlyingpercep-
tualfactors on theformation andreference oflinguistic Certaincolors
categories.
appearto be universallysalient.There are also universaisin some aspectsof
colornaming.How (bywhatmechanism)mightthesaliencybe relatedtothe
naming?Whatwe areaskingforis an accountofthedevelopment (bothin the
sense of individuallearningand the evolutionof languages)of color names
whichwill specifytheprecisenatureoftheroleplayedbyfocalcolorsin that
development.
Such an accountis relatedtoissuesmoregeneralthanthatoflinguisticrela-
tivityalone. Learningtheoriesin psychologytendto be designedprimarily
to account forconnectionsformedbetweeninitiallyarbitrarystimuliand
responses.The conceptslearnedin typicalconcept-formation tasks(37X38)
arealsoarbitrary; whata subjectlearnswhenhe learnssucha conceptis a clear-
cut rule,usuallystatedin termsofcombinationsofthediscreteattributes of
artificialstimuli, which define the boundaries of membershipin the
experimenter-determined "positivesubset"(forexample,"anything whichis
squareand has twobordersaroundit" is a memberofthe"concept").Color
categories,however, appeartobe conceptswitha verydifferent kindofstructure.
Rosch(39) proposedthe followingaccount of the developmentof color
names:thereareperceptually salientcolorswhichmorereadilyattractatten-
tion,evenofyoungchildren,(40) and aremoreeasilyremembered thanother
colors.When categorynamesarelearned,theytendto become attachedfirst
to the salient stimuli(only latergeneralizingto other,physicallysimilar,
instances), and bythismeansthese"naturalprototype" colorsbecomethefoci
oforganization forcategories. How can thisaccountbe tested?In thefirstplace,
itimpliesthatitis easiertolearnnamesforfocalthanfornon-focal colors.That
is,notonlyshouldfocalcolorsbe moreeasilyretainedthannon-focal in recog-
nitionovershortintervals (as has alreadybeen demonstrated), buttheyshould
also be morereadilyrememberedin conjunctionwithnames in long-term
memory.In thesecondplace,sincea colorcategoryis learnedfirstas a single
namedfocalcolorand secondas thatfocalcolorplus otherphysicallysimilar
colors,colorcategories in whichfocalcolorsarephysically centralstimuli("cen-
tral"in termsofsomephysicalattribute, suchas wavelength)shouldbe easier
to learnthancategoriesstructured in some othermanner(forexample,focal
colorsphysically peripheral, or internominal colorscentral,and no focalcolors
at all).
A testofthesehypothesesobviouslycould notbe performed withsubjects
who already knew a setof basic chromatic color terms providedbytheirlan-
guage.This bringsus toanotherimportant possiblemethod forcross-cultural
research whichhas seldombeenapplied-a learningparadigm.Manycultures
lackcodes (or a fullelaborationofcodes) forsome domain.If an investigator
has theoriesaboutthatdomain,insteadofframing his hypothesesin termsof

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
272 Et cetera • FALL1987

deficitsin performance resulting fromthelackofcodes (withattendant prob-


lems in interpreting absolute differences betweencultures),he can frame
hypothesesin termsoflearning thecodes forthatdomain.Codes can thenbe
taught-theinput stimuli preciselyspecifiedand controlledwithinthecon-
textoftheexperimentin accordancewiththe relevanthypotheses.Since the
variationsarewithinculture,anygeneraldifficulty whichthepeoplemayhave
withthelearningtaskper se will not influencethe conclusions.The Dani,
withtheirtwo-term colorlanguage,providedan idealopportunity toteachcolor
names.
Three basic typesofcolorcategoryweretaught.In Type 1, thephysically
central(i.e., ofintermediate value in wavelengthor brightness) chip ofeach
category was the fixaicolor, and the flanking chipswere drawn from theperiph-
ery ofthat basic name area. In Type 2, central chipslay in the internominal
areasbetweenBerlinand Kay's best-example clusters;flankingchips,thus,
tendedtobe drawnfromthebasic color-nameareason eitherside. Since two
different basic color-namechipswereincludedin the same Type 2 categor-
ies,these categories"violated"thepresumednaturalorganization ofthecolor
3
space. Type categories were located in the same as
spaces Type 1; however,
insteadofoccupyinga centralposition,thefocalcolorwas nowtoone side or
the otherof the three-chip category.
Subjectslearnedthecolornamesas a paired-associate task,a standard learning
taskinwhichsubjectslearntogivea specificresponsetoeachofa listofstimuli.
In thepresentcase,colorswerethestimuli,and thesame Dani wordwas the
correctresponseforthethreecolorsin a category.Findingsuitable"names"
forthecolorsat firstseemeda seriousobstacleto thestudysinceDani would
notlearnnonsensewords,eventhoseconstructed accordingtotherulesofthe
Dani language.Here is an exampleofa case in whichitwas necessarytomake
thetaskculturallymeaningfulifit was to be performed at all. Eventually,it
was foundthattherewas a setofkingroupscalled sibs(somethinglikeclans)
whosenameswereall wellknowntotheDani and whichtheDani couldreadily
learnas names forthe color categories.(Sibs did not have particularcolors
associatedwiththemin Dani culture.)The taskwas describedto each sub-
jectas learninga newlanguagewhichtheexperimenter wouldteachhim.The
subject was told the"names" for all of the color chips,then presented witheach
chip and to
required respond with a name. Chips were shown in a different
randomordereachrun,fiverunsa day,withfeedbackaftereachresponse, until
the criterionof one perfectrunwas achieved.
The resultsofthe learningsupportedRosch'saccountof the roleof focal
colorsin thelearningofcolornames.In thefirstplace,thefocalcolorswere
learnedwithfewererrorsthanothercolors,evenwhentheywereperipheral
membersofthecategories. In thesecondplace,theType 1 categories in which
focalcolorswerephysicallycentralwerelearnedas a setfasterthaneitherof
theothertypes.The Type2 categories, whichviolatedthepresumednatural
organization ofthe color space, were the mostdifficult ofall to learn.Thus,

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
Linguistic Relativity 273

theidea ofperceptually salientfixaicolorsas "naturalprototypes" (ratherlike


Platonicforms)forthedevelopment and learningofcolornameswassupported.
We havebeen speakingoffixaicolorsas "perceptually" salient.Is thisjust
a metaphor,or is therean actualmechanismof colorvisionwhichcould be
responsibleforthesalience?The answeris "both."There is a theoryofcolor
perception(41) supported by both psychophysical and physiological
data(42X43)whichclaimsthattheprimary colorsred,green,yellow,and blue
correspond to physiologically"uniquehues."To getsomenotionofthemeaning
ofunique hues, imaginethattherearetwo"opponent"color-codingsystems
in theprimatenervoussystem(in actualfeet,in the lateralgeniculate),each
ofwhichcan respondpositively or negatively.One systemis responsive to red
and greenwave lengthsof light,the otherto yellowand blue wavelengths.
Thinkoftheprobabilities ofstimulation ofeach systemdistributed overwave
lengths.Therewillbe fourpoints(particular wavelengths)at whichone system
respondsuniquely;thatis, a pointat whichtheyellow-bluesystemis neutral
and thered-green systempositive,a pointat whichyellow-blue is neutraland
red-green negative,and pointsat whichred-green is neutraland yellow-blue
positiveand negative.
Do thewavelengthsoftheproposedfourunique hue pointscorrespondto
"fixai"colors?Theycannotcorrespondexactlybecausephysiological and psy-
chophysicalvisualresearchtendsto be performed withmonochromatic light
(radiantlightofa singlewavelength),whereasMunsellchipsare"broadband"
light(reflectedlightcontainingmanywavelengths).However,thedominant
wavelengthofeach Munsell chip has been calculated.(44) It is, in fact,the
case thatthedominantwavelengthsoffocalred,yellow,green,and blue cor-
respondreasonablywell to the proposedunique hue points.Evidenceof an
evenmoredirectmatchoffocalyellow,green,and blue to unique hue points
(redwas nottested)is providedin McDaniel. (45) While unique hue points
arenotpresently an unchallengedphysiologicaltheory, and whilethetheory
failstoaccountfortheotherfourproposedbasic chromaticcolorterms(pink,
orange,brown,and purple),itdoes lendconsiderableconcreteness tothesup-
positionthatfixaicolorsarephysiologically, ratherthanmysteriously, salient.
At thispoint,thereadermaywell feela sense ofdiscontent.We appearto
haveconcludedthatcolorterminology is entirelyuniversal.But whatofcolor
termboundaries,and what of the degreeof elaborationof secondarycolor
terms?If colortermsmakeno difference to perception,cognitiveprocesses,
communication, orlife,whyshouldlanguageshaveanycolortermsat all,much
less differencesin terms?
What are colortermsused for?One theoryis thatwe have themin order
tocommunicateaboutobjectswhicharethesame exceptforcolor.All ofthe
cultureswhichhavefewerthanthefullcomplementof 11 basictermsarealso
technologically notat an industriallevel.Accordingtothistheory, colorterms
onlybecome necessaryforcommunicationwhen manufactured objectscan
be producedin multitudes, and coloringagentsareavailableforimparting dif-

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
274 Et cetera • FALL 1987

ferentcolorsto theotherwiseidenticalobjects.A paradigmaticsituationfor


using color termsin thiscontextwouldbe tosay,"Bringme theorangebowl,"
therebyspecifying whichof several,otherwiseindistinguishable, bowlswas
desired.
But whyshouldanyonewanttospecifythe"orangebowl"?Thinkaboutthe
contextsin whichyouactuallypayattention tosubtledifferences in color.They
areprobablyactivities suchas decidingwhatarticlesofclothingtowearsimul-
taneously,decoratinghouses, landscaping gardens,and producingand
appreciatingworksof art.
There is one studywhichbearson thispoint.Greenfieldand Childs stud-
ied theeffectofknowinghowto weavecertainpatternsin clothupon pattern
conceptionamongtheZinancantecosofChiapas, Mexico. (46) The patterns
consistedof simplegroupsof redand whitethreads.Subjectswereaskedto
"copy"thepatternbyplacingsticksintoa frame.Theyweregiventheirchoice
ofvariouswidthsand colorsofsticks.While some subjectsused onlythered
and whitesticksto copytheredand whitepatterns,othersfreelysubstituted
pinkforwhiteand orangeforred.A separatetestdeterminedthatall subjects
coulddiscriminate thedifferences betweenred,orange,pink,and whitesticks
equally well. The important pointofourargumentis thatittendedtobe sub-
jectswho named the red,pink,orange,and whitestickswithdifferent names
who adheredstrictly totheredand whitesticksforcopyingthepatterns;sub-
jectswho used onlya singletermforwhiteand pinkand a singletermforred
and orangeweretheones who tendedto makethesubstitutions. It maywell
be thatit is in little-understood domainssuch as aestheticjudgmentthatthe
use ofcolortermswillbe foundto"makea difference." (Of course,theZinan-
cantecoswho used differentiating termsmayhavedone so becausetheywere
themoresensitivetoaestheticdifferences.) Whatdifference termsdo makecan
now be exploredagainstour backgroundof knowledgeof whatis universal
in color.
Webeganwiththeidea ofcoloras theidealdomainin whichtodemonstrate
theeffects ofthelexiconofa languageon cognition,thereby supporting a posi-
tionoflinguisticdeterminism. Instead,we have found that basic color termi-
nologyappears to be universaland thatperceptuallysalientfocalcolors appear
toformnaturalprototypes forthedevelopment ofcolorterms.Contrary toinitial
ideas,thecolorspace appearstobe a primeexampleoftheinfluenceofunder-
lyingperceptual-cognitive factorson linguisticcategories.

Other Natural Categories


Is color the only domain structuredinto "naturalcategories"?It seems
unlikely.Color may,in feet,providea bettermodel forthe natureofhuman
categorizingthando theartificialconceptsused bypsychologistsin concept-
formation research.In thefirstplace,thereare otherdomainsin whichper-
ceptuallysalientnaturalprototypes appearto determinecategories;geomet-
ofemotionarecasesin point.In thesecondplace,
ricformsand fecialexpression

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
Linguistic Relativity 275

categoriesnotbased on biologically"given"prototypes mayalso obeypsycho-


logicallawsfortheperception andsegmentation ofexperience,thereby yielding
naturallystructuredcategories.
That thereis something particularly"wellformed" aboutcertainforms, such
as circlesand squares,waslongagoproposedbytheGestaltpsychologists. Rosch
testedthehypothesisthatsuch formsact as naturalprototypes in theforma-
tionofformcategories justas focalcolorsdo forcolorcategories.(47) The Dani
alsodo nothavea terminology fortwo-dimensional geometric forms, and some
pilotstudiesshowedthattheyneitherpossessedusable circumlocutions for
referring to formsin a communicationtask,nor did theytendto sortforms
byformtype.Thus, itwas reasonabletoteachDani formconceptsjustas they
had been taughtcolorconcepts.The logic of the form-learning experiment
was thesame as thatofthecolor-learning. Circle,square,and equilateraltri-
angleweretakenas thepresumednaturalprototypes ofthreeformcategories.
In the"naturallystructured" categories,these"good forms"werephysically
centraltoa setofdistortions (suchas gapsintheformorlineschangedtocurves).
In othercategories,a distortedformwas thecentralmember,thegood forms
peripheral.The resultsmirrored thoseforcolor.The good formsthemselves
werelearnedfesterthanthedistortedforms,and thesets of formsin which
thegood formswerecentralwerelearnedfesterthansetsin whichtheywere
peripheral.Furthermore, fortheforms(thoughnotforthecolors),Dani were
willing,at the conclusionof learning,to pointto whichstimulustheycon-
sideredthebestexampleofthenametheyhad justlearned.The good forms
tendedto be designatedas thebest examples,evenwhen theywereactually
peripheral totheset;itwas as thoughsubjectsweretrying to structure thecat-
egoriesaroundthegoodforms, evenwhentheactualsetswerestructured other-
wise.
Facial expressionsofemotionarea surprisingadditionto theclass ofnatu-
ralcategories. Not onlyweretheyonce notconsidereduniversal,buttherewas
considerabledoubt that,evenwithinone culture,emotioncould be judged
betterthanchancefromthehumanfece.(48) As hadbeenthecasewithcolors,
such judgmentsseemedto stemfromtheunsystematic employmentofmis-
cellaneousfacialexpressionsin judgmentexperiments. Ekman claimedthat
therearesixbasic humanemotions(happiness,sadness,anger,fear,surprise,
and disgust)and thateach is associatedwitha quite limitedrangeof fecial
musclemovements constituting a pureexpressionofthatemotion;otherexpres-
sionstendto be blendsofemotions,or ambiguousor nonemotionalexpres-
sionswhichcould notbe expectedto receivereliablejudgments.(49) When
Ekmanputtogether setsofpicturesofpureexpressionsoftheproposedbasic
emotions,he foundthatthesepictureswerejudgedcorrectly by Americans,
Japanese,Brazilians,Chileans,and Argentinians. Furthermore, twopreliter-
ate New Guinea groupswithminimalcontactwithCaucasian fecialexpres-
sion,theForeand theDani, wereable todistinguish whichoftheexpressions
was meanton thebasis of storiesembodyingthe appropriateemotion.Like

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
276 Et cetera • FALL1987

color,universality was discoveredin facialexpressionsofemotiononlywhen


an investigator thoughtto ask, not about all possiblestimuli,but about the
prototypes (bestexamples)ofcategories.As is thecase forcolorterms,there
appearsto be a residualfunctionofemotionnamesthemselves.In a commu-
nicationtask in which one subject attemptedto communicateverballyto
anotherwhichone setofpicturesoffeceswas intended,Americansperformed
farbetterwithpicturesofthepureemotionsthanwithambiguousexpressions;
Dani, however, who lacka setofemotionterms,showedno difference in per-
formancebetweenthetwotypesof pictures.(50)
It is unreasonableto expectthathumanscome equipped withnaturalpro-
totypesin all domains.Dogs, vegetables,and Volkswagens, forexample,are
probablyculturally relative. Yet such categoriesmay also possessan "internal
structure" whichrendersthemmoresimilartocolorthantoartificial categories.
That is, thecolor,form,and emotionalexpressioncategorieswerecomposed
of a "coremeaning"(the clearestcases, best examples)of the category, "sur-
rounded"byothercategory membersofdecreasingsimilarity tothecoremean-
ing.Thinkaboutthecommonsemanticcategory "dog."Whichis a betterexam-
of
ple your idea orimage of what that word means (whichis doggier?):a German
shepherdor a dachshund?Rosch had college studentsratemembersof a
numberofsemanticcategories as totheirprototypicality and foundhighagree-
mentin judgmentbetweensubjects.(51) Evidencehas sincebeen obtained,
in a varietyoftasks,thatsuch categoriesseem to be "storedmentally," notas
a listoflogicalcriteriaforcategorymembership, but ratherseemtobe coded
in a "shorthand" formconsistingofa fairly concreterepresentation ofthepro-
totype.(52X53)
Ifinternal structure and prototypes, whether "given"orlearned,areimportant
aspects in the and
learning processing ofsemantic categories,thefeethasimpli-
cationsforcross-cultural research. Presentanthropological linguistic
techniques
(forexample,componentialanalysis)tendto emphasizediscoveryofthemin-
imal and mostelegant,logicalcriterianeededto determinemembershipin,
and distinctions between,classes.Analysisofthebest-example prototypes of
categoriesmayprovide us with a new,psychologically real,and fruitfulbasis
forcomparisonof categoriesacrosscultures.
Evencompletely asidefrominternal structure,givenanycollectionofstimuli
or culturalenvironment, it is unreasonableto expectthatcategorieswill be
formedrandomly. For example,thereareundoubtedlypsychological rulesfor
perceiving "clusters" of stimuli and "gaps" between stimuli. Such factorsas
of
frequency particularobjects, order of encounter with the objects,"density"
ofnonidentical butsimilarstimuli,andtheextenttowhichobjectsin one"clus-
ter"aredistinctively different fromobjectsin other"clusters"areexamplesof
thekindsoffactors whichmightdetermine psychological grouping.Of course,
categories of all not
typesprobably only have labels,but also havesomerationale
whichmakesthemnot purelyarbitrary but, rather,natural categories.

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
Linguistic Relativity 277

CONCLUSIONS
We began withthenotionoflinguisticrelativity definedin termsofinsur-
mountabledifferences in theworldviewsofculturesbroughtaboutbydiffer-
encesin naturallanguages.Because ofthevarietyofrequirements forspeci-
ficityand cross-culturalcontrolsin testingsuch assertions,we werereduced
totheferless sweepingclaimthatcolornamesaffect someaspectsofthought.
However,we discoveredthatcolorsappearedtobe a domainsuitedto demon-
stratejusttheoppositeoflinguisticrelativity, namely,theeffect ofthehuman
perceptualsystemin determining linguisticcategories.Verysimilarevidence
existsin thedomainsofgeometric formand emotioncategories. Furthermore,
psychologicalprinciplesofcategorization mayapplyto the formation of all
categories,even in culturallyrelativedomains.
At present,theWhorfianhypothesis notonlydoes notappeartobe empir-
icallytruein anymajorrespect,but it no longerevenseemsprofoundly and
ineffably true.Whyhas itbeen so difficult todemonstrate effectsoflanguage
on thought? Whorfreferred tolanguageas an instrument which"dissects"and
categorizes "nature."
In thefirstpartofthechapter,we sawthatithas notbeen
establishedthatthecategorizations providedbythegrammarofthelanguage
actuallycorrespondto meaningful cognitiveunits.Fromthelatterpartofthe
chapter,we can nowsee thatforthevocabularyoflanguage,in and ofitself,
tobe a molderofthought, lexicaldissections
and categorizationsofnaturewould
havetobe almostaccidentally formed, ratheras thoughsomeJohnny Appleseed
hadscattered namedcategories capriciouslyovertheearth.In 6a, the"effects"
ofmostlexicallinguisticcategoriesareprobablyinseparablefromtheeffects
ofthefactors whichled initiallytotheformation and structuringofjustthose
categoriesratherthansome others.It would seem a farrichertaskforfuture
researchto investigate the entirecomplex of how languages,cultures,and
individualscome,in thefirstplace,to"dissect," and "name"nature
"categorize,"
in thevariouswaysthattheydo.

NOTES AND REFERENCES


1. B.L.Whorf, Language, andreality
thought : Selected ofBenjamin
writings LeeWhorf,ed.
J.B.Carroll (Cambridge,Mass.:MIT Press,1956),212-213.
2. Whorf, Language, thoughtandreality,59.
3. cf.C.C.Fries,Thestructure Anintroduction
ofEnglish: totheconstruction
ofEnglishsen-
tences
(NewYork:Harcourt, Brace,1952).
4. Roger W.Brown, Words andthings(NewYork:' FreePress,1958).
5. S. Ervin,
"Theconnotations ofgender,Word, 18(1962):248-261.
6. Thisiscalledthesemantic differential
-cf.C.E. Osgood,G.J.Suci,andP.H.Tannen-
baum,TheMeasurement ofmeaning (Urbana:UniversityofIllinoisPress,1957).
7. Whorf, Language, thoughtandreality
, 243.
8. M. Black,"Linguistic Theviews
relativity: ofBenjamin LeeWhorf?' Review
Philosophical
68(1959):252-253.
9. J.B.CarrollandJ.B.Casagrande, "Thefunction oflanguage inbehavior,"
classifications
inReadings insocial 3rded.,E.E.Maccoby,
psychology, T.M.Newcomb, andE.L. Hart-
ley,eds.(NewYork: Holt,Rinehart & Winston, 1958).

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
278 Et cetera • Fall 1987

10. Fora review ofstudies, seeR. Serpell, "Theinfluence oflanguage,education andcul-


tureonattentional preferencesbetween colour andform," InternationalJournal ofPsy-
chology4 (1969):183-194.
11. Seelatter
partofthisarticle aswellasB.Berlin andP.Kay,Basiccolor terms: Their univer-
andevolution
sality (Berkeley: University ofCalifornia Press,1969).
12. E.R.HeiderandD.C.Oliver, "Thestructure ofthecolorspaceinnaming andmemory
fortwolanguages," Cognitive Psychology 3 (1972):337-354.
13. Z.M.Istomina, "Perception andnaming ofcolorinearly childhood,"Soviet Psychology
andPsychiatry 1 (1963):37-46.
14. E.V.Clark,"What's ina word? Onthechild'sacquisition ofsemantics inhisfirst lan-
guage,"inCognitive development andtheacquisition oflanguage,T.E.Moore, ed.(New
York:Academic Press,1973).
15. C.N.Cofer andD.R.Bruce, "Form-class asthebasisofclustering intherecall ofnonas-
sociatedwords," Journal ofVerbal Learning andVerbal Behavior 4 (1965):386-389.
16. Fora discussion ofsomeofthecomplex issuesinvolved inrecallofcategorized lists,see
C.N.Cofer, "Onsomefactors intheorganizational characteristics
offree recall," Ameri-
canPsychologist 20(1965):261-272.
17. D.D.Wickens, "Encoding categories ofwords: Anempirical approach tomeaning," Psy-
chologicalReview 77(1970):1-15.
18. Therearetasks inthehuman learning andmemory inwhich
literature grammatical class
doesnothaveaneffect; however, thesearealltasks inwhich predicted differences are
derived from andeasilyexplained bytheformal syntax ofgrammatical class-thatis,
from privileges ofoccurrence instrings ofwords.
19. G.A.Miller, "Themagical number seven,plusorminus two,"Psychological Review 63
(1956):81-97.
20. Fora more detailed discussion oftheepidemiological model,seeR.A.LeVine,"Cross-
culturalstudy inchildpsychology," inCarmichaeVs manual ofchild
psychology, 3rded.,
P.H.Müssen, ed.(NewYork: Wiley, 1970).
21. Foraninteresting example ofitsuse,seeM.H.Segal, D.T.Campbell, andM.J.Herskovitz,
Theinfluence ofculture onvisual perception (Indianapolis: Bobbs-Merrill, 1966).
22. Ananthropologist's report oftheexistence, nonexistence, ordenotation ofsemantic cat-
egoriesinthelanguage ofa culture hehasstudied isusuallyaccepted atfacevalue.
Although wewillfollow thatpractice here, itshouldberealized thatthere areaddi-
tionalmethodological problems withtraditional fieldtechniquesforstudying language.
Fora discussion, seeE.R.Heider, "Probabilities,sampling, andethnographic method:
ThecaseofDanicolornames," Man7 (1972):448-466.
23. Colorplatesillustrating whatthesedimensions actuallylooklikeareprinted inmany
standard introductory psychology textbooks -forexample, inRuch'sPsychology and
Life(6thEdition) andinHilgard, Atkinson, andBower's Introduction toPsychology
(5thEdition).
24. E.H. Lenneberg, Biological foundations oflanguage (NewYork: Wiley, 1967).
25. R.L. De ValoisandG.H. Jacobs, "Primate colorvision," Science162(1968):533-540.
" inlanguage
26. R.W.Brown andE.H. Lenneberg, study andcognition," Journal ofAbnor-
malandSocialPsychology 49(1954):454-462.
27. G.K. Zipf,Thepsycho-biology oflanguage (Boston: Houghton-Mifflin, 1935).
28. Infact,another variable, "communication accuracy," wasfound tocorrelate with memory
moregenerally thancodability. (SeeD. LantzandV.Stefflre, "Language andcogni-
tionrevisited,"Journal ofAbnormal andSocialPsychology 69(1964):472-481 .)(Seealso
V.Stefflre,V.Castillo Vales, andL. Morley, "Language andcognition inYucatan: A
cross-cultural replication," Journal ofPersonality andSocialPsychology 4 (1966):112-
115.)However, because thislineofresearch is morerelevant totherelation between
interpersonal andintrapersonal communication thanitistotherelation between a lin-
guisticdomain andthenonlinguistic domain whichitencodes (seeLenneberg, Bio-

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions
Linguistic Relativity 279

foundations
logical oflanguage), itwillnotbepursued further here.
29. Berlin andKay,Basiccolor terms.
30. Munsell ColorCompany, Dominant wavelength andexcitation purityfordesignatedMunsell
colornotation (Baltimore: Munsell ColorCompany, 1970).
31. E.R. Heider, "Universais incolornaming andmemory," Journal ofExperimental Psy-
chology93 (1972): 10-20.
32. Thesemaynotactually bethe"best" chipstorepresent focalandnon-focal colors.Nei-
therBerlinandKay'slinguistics northeir research methods areabovereproach (see
N.RHickerson, "Review of'Basiccolor terms:Theiruniversality andevolution,'"Inter-
nationalJournal ofAmerican Linguistics 37(1971):257-270). Berlin andKaymayhave
included somecolors intheir basicnamelistwhich should beconsidered secondary
names, ormayhaveassigned secondary status tolegitimate basicterms; orthey may
havesystematically skewed thelocation oftheirbest-example clusters bytheuseofbilin-
gualsas subjects. Allsuch"errors" wouldonlycontribute to"noise"inthepresent
author'sresearch design andmakeitmoredifficult todemonstrate differ-
significant
encesbetween focalandnon-focal colors.
33. K.G.Heider, TheDugum Dani: A Papuanculture intheHighlands ofWest NewGuinea
(Chicago: Aldine, 1970).
34. Heider, "Probabilities, sampling, andethnographic method."
35. Heider andOliver, "Thestructure ofthecolorspaceinnaming andmemory fortwolan-
guages."
36. Heider, "Universals incolornaming andmemory."
37. cf.L.E. Bourne, Humanconceptual behavior (Boston: Allyn& Bacon,1968).
38. cf.J.S.Bruner, J.J. Goodnow, andG.A.Austin, A study ofthinking (NewYork: Wiley,
1956).
39. E. Rosch, "Ontheinternal structure ofperceptual andsemantic categories,"inCogni-
tiveDevelopment andtheacquisition oflanguage (NewYork:Academic Press,1973).
40. E.R. Heider, "'Focal'colorareasandthedevelopment ofcolornames," Developmental
Psychology 4 (1971):447-455.
41. E. Hering, Outlines ofa theory ofthelight sense
, trans. L.M. Hurvich andD. Jameson
(Cambridge: Harvard University Press,1964).
42. De ValoisandJacobs, "Primate colorvision."
43. Hering, Outlines ofa theory ofthelight sense.
44. Munsell ColorCompany, TheMunsell bookofcolor.
45. C.K. McDaniel, "Hueperception andhuenaming" (Unpublished honorsthesis,Har-
vardCollege, April1972).
46. P.M.Greenfield andC. Childs, "Weaving colorterms,
skill, andpattern representation
among theZinacantecos ofSouthern Mexico": Adevelopmental study(Unpublished
manuscript, Center forCognitive Studies, Harvard University, 1971).
47. E. Rosch, "Ontheinternal structure ofperceptual andsemantic categories."
48. J.S.Bruner andR. Tagiuri, "Theperception ofpeople," inHandbook ofsocial
psychol-
ogy,Vol.2,ed.G. Lindzey (Cambridge: Addison-Wesley, 1954).
49. P. Ekman, "Universal andcultural differences in facialexpressions ofemotion, in
NebraskaSymposium onMotivation, ed.J.Cole(Lincoln: University ofNebraska Press,
1972).
50. E. Rosch,"Verbal andnonverbal communication ofthesamearray: ofcode,cul-
Effects
ture,andclass(Unpublished manuscript, 1973).
51. E. Rosch,"Ontheinternal structure ofperceptual andsemantic categories."
52. E.R.Heider, "Nature ofthemental codefornatural categories" (Paperpresented atthe
meeting ofthePsvchonomics Society, St.Louis,November, 1972).
53. E. Rosch,"Ontheinternal structure ofperceptual andsemantic categories."

This content downloaded from 198.91.37.2 on Sun, 27 Dec 2015 14:23:45 UTC
All use subject to JSTOR Terms and Conditions

You might also like